I am evaluating the impact of a project with NON-RANDOMIZED multiple interventions (3 treatment groups and 1 control/comparison group). I have a household survey data for two periods (baseline and follow-up).
To deal with selection bias (in particular), I am considering the use of Stata's teffects ipwra to obtain the average treatment effects.

I want to know from the experts in impact evaluation which estimation procedure is most appropriate.. and also if teffects ipwra does the same job as diff-in-diff regression with inverse probability weighting based on GPS?


Thank you